Simple linear regression with Gretl (no programming required)

[Charles Durfee]

Gretl is a free open source statistical analysis tool, similar to SAS. You can download it here.

You can download the dataset here.

Step 1 Open file:
https://www.kaggle.com/rohankayan/years-of-experience-and-salary-dataset/version/1
File> Open Data> User File> Select Salary_Data.csv (you may need to change the drop down to “Select all files”)
Click “No” when asked “Do you want to give the data a time-series or panel interpretation)

Step 2 Run model:
Model> Ordinary Least Squares
Select “Salary” as Dependent variable
Select “YearsExperience” as Regressors
Click “Ok”
Note smaller the p-value the better: must be lower than 0.05 to be significant.
Note under the coefficient heading: const is the starting salary with no experience, YearsExperience is the amount added for each year(slope of the line).
Note the R-Squared value closer to 1 is better

Step 3 Graph Model:
Analysis> Forecast
Note the blue line is the models prediction for the salary
Note there is a 95% chance that the salary will be between the green lines

[Author]

Posts